Lies, damned lies, statistics – and Scottish samples in opinion polls


by Dave Taylor

Last week we had a set of polls across the UK on Westminster voting intention, the results contained interesting Scottish sub-samples.

Now, everyone knows (or should know) that sub samples are wholly unreliable because they aren’t balanced to represent the demography of Scotland, and the numbers polled are so small that the margin of error is massive.

But, subsamples notwithstanding, it seemed unlikely that all 4 pollsters should all have errors in the same direction – towards the SNP.

Those polls showed the following pattern:

Party ICM ComRes Populus MORI 4 poll average
SNP 53% 42% 39% 39% 42%
Lab 29% 35% 37% 31% 34%
Con 9% 18% 11% 12% 14%
L_D 9% 4% 5% 8% 6%
Oth 0% 1% 8% 9% 4%

However, at the same time as these results were being recorded, another pollster was showing something completely different.

YouGov (which polls a selection of its internet panel) showed not an SNP lead, but a rather big Labour lead over the SNP – 42% to 28%.

This week we have had further GB polling, with the following from the Scottish samples –

ComRes (using telephone polling) has SNP and Lab both on 33%, Con on 12% and LD 2%.

YouGov has Lab 46% : SNP 25% : Con 16% : LD 9%

Angus Reid (using a similar type of internet panel to YouGov) has Lab on 36% : SNP 29% : Con 12% : LD 2%

For those who take a keen interest in polling this suggests that there may be a systemic difference between pollsters using internet panels as opposed to results using more traditional methods.  Interestingly, traditional methods were better at coming somewhere near predicting the Scottish General Election.

In August, Ipsos-MORI will publish their quarterly Scottish poll that will probably be the first real indicator of Scottish opinion of Westminster voting intention.

If that shows that Labour has a big lead over the SNP for Westminster, then it would confirm the “conventional” set of assumptions about Scottish politics, namely that people opt for Labour at Westminster and the SNP at Holyrood.

If the two big Scottish parties are neck and neck, then that would suggest a fairly important shift in Scotland and a very real concern for not just ‘Scottish’ Labour but the whole UK party.